A one-layer projection neural network for linear assignment problem

This paper presents an improved one-layer projection neural network for solving the linear assignment problem. The assignment problem is first converted into a linear programming problem, then a corresponding recurrent neural network is constructed. The optimality and global convergence of the proposed neural network are analysed and proved. Compared with existing neural networks for linear assignment problem, the proposed model does not include any design parameter and the activation function is continuous, which is more convenient for real application. Simulation results on two numerical examples are discussed to demonstrate the effectiveness and characteristics of the proposed neural network.